This commit is contained in:
frostydad 2022-10-01 15:32:18 -06:00
parent d6b9e9ddd9
commit 4a5e84afff
8 changed files with 122 additions and 145 deletions

View File

@ -4,25 +4,31 @@ _Google Colab notebooks ported to standard Jupiter for Paperspace._
Don't post bug reports on the Kazakhstan rock collection forum. Create an issue here.
### Getting Started
[Docs/Paperspace Guide for Idiots.md](https://github.com/Engineer-of-Stuff/stable-diffusion-paperspace/blob/main/Docs/Paperspace%20Guide%20for%20Retards.md)
### Notebook Descriptions
**StableDiffusionUI_Voldemort_paperspace.ipynb**
#### StableDiffusionUI_Voldemort_paperspace.ipynb
Voldemort's webUI. His repository is updated frequently so the notebook will update the local copy if it's already installed. Can run the normal model or waifu-diffusion.
AUTOMATIC1111's webUI. Designed to get a newbie set up, features lots of error checking and automation.
#### Cyberes_Textual_Inversion_Training.ipynb
All the commands you need to do create a textual inversion embedding. Does not hold your hand or guide you through the process. Any issues related to this notebook will be closed.
**CodeFormer_Inference_Simplified.ipynb**
#### lfs/latent-diffusion
CodeFormer face fixer standalone. Can process images in bulk.
Holds stuff for latent-diffusion. Probably not needed but latent-diffusion had them stored on some random file hosting service.
#### /other Directory
Other notebooks and code.
**stable_diffusion_webui_hlky_paperspace_09-07-2022.ipynb**
hlky's webui
- CodeFormer_Inference_Simplified.ipynb
- stable_diffusion_webui_hlky_paperspace_09-07-2022.ipynb
- sd-concepts-library_Exporter.ipynb
- Huggingface Textual Inversion Training.ipynb
- download_sd-concepts-library.py
- Misc. old stuff

View File

@ -0,0 +1,77 @@
# Getting Started on Paperspace for Retards <3
![paperspace](https://raw.githubusercontent.com/Engineer-of-Stuff/stable-diffusion-paperspace/main/docs/assets/paperspace.png)
**What is Paperspace?**
Basically, a cloud compute service for AI development. You run your code on their hardware.
**Why Paperspace?**
Paperspace is convenient, free, and you probably don't have a computer good enough to run the AI. They're aimed at professionals whereas Colab is literally for children.
You'll run your code on their powerful GPUs which is much faster and easier than locally on your computer.
**Do I have to pay?**
Only if you feel limited by their free tier. You should have everything you need to generate ~~porn~~ really neat images. I'm pretty Paperspace is cheaper than Google Colab Pro.
**Why not Google Colab?**
Google Colab is an alternative, but their free tier is more restrictive than Paperspace's and has less powerful hardware. And, knowing Google, they probably record what you generate.
**But I want to run it on Google Colab!**
Ok fine. [Here's the official colab notebook by Voldy himself.](https://colab.research.google.com/drive/1Iy-xW9t1-OQWhb0hNxueGij8phCyluOh)
**But I'd rather run it on my own computer!**
Running it in the cloud is much easier and you're less likely to mess up your OS, but since you insist here's a Docker container to make is simpler: [AbdBarho/stable-diffusion-webui-docker](https://github.com/AbdBarho/stable-diffusion-webui-docker)
## Let's do it!
1. [Create an Account](https://console.paperspace.com/signup) You will be asked for your phone number. **You can use a VOIP number such as Google Voice, they don't block it!**
2. Click this button and fill out the form that pops up
![create_project](https://raw.githubusercontent.com/Engineer-of-Stuff/stable-diffusion-paperspace/main/docs/assets/create_project.png)
3. Click this button to create a notebook
![create_notebook](https://raw.githubusercontent.com/Engineer-of-Stuff/stable-diffusion-paperspace/main/docs/assets/create_notebook.png)
4. Under "Runtime" select "Start from Stratch"
![start_scratch.png](https://raw.githubusercontent.com/Engineer-of-Stuff/stable-diffusion-paperspace/main/docs/assets/start_scratch.png)
5. Download [StableDiffusionUI_Voldemort_paperspace.ipynb](https://github.com/Engineer-of-Stuff/stable-diffusion-paperspace/blob/main/StableDiffusionUI_Voldemort_paperspace.ipynb)
6. Upload that file to your notebook
![upload](https://raw.githubusercontent.com/Engineer-of-Stuff/stable-diffusion-paperspace/main/docs/assets/upload.png)
7. Follow directions
![final](https://raw.githubusercontent.com/Engineer-of-Stuff/stable-diffusion-paperspace/main/docs/assets/final.png)
## Launch the Jupyter Notebook
Click this symbol in the left vertical menubar.
![open-jupiter](https://raw.githubusercontent.com/Engineer-of-Stuff/stable-diffusion-paperspace/main/docs/assets/open-jupiter.png)
You can access the terminal in the Jupyter Notebook.
## I need more help :(
Here's some other guides.
- [FINAL GUI RETARD GUIDE](https://rentry.org/voldy)
- [Using the WebUI](https://rentry.org/voldy)
- [Using the Inpainter](https://rentry.org/drfar)
- [Textual Inversion](https://rentry.org/aikgx)
- [Crowd-Sourced Prompts](https://lexica.art/)
- [Artist Name Prompts](https://sgreens.notion.site/sgreens/4ca6f4e229e24da6845b6d49e6b08ae7?v=fdf861d1c65d456e98904fe3f3670bd3)
[Images hosted on Github](https://github.com/Engineer-of-Stuff/stable-diffusion-paperspace)
[Github Mirror](https://github.com/Engineer-of-Stuff/stable-diffusion-paperspace/blob/main/docs/archives/Getting%20Started%20on%20Paperspace.pdf)

31
other/README.md Normal file
View File

@ -0,0 +1,31 @@
# Other
#### CodeFormer_Inference_Simplified.ipynb
CodeFormer face fixer standalone. Can process images in bulk.
#### stable_diffusion_webui_hlky_paperspace_09-07-2022.ipynb
An alternative to AUTOMATIC1111's WebUI from a guy named hlky. Ported to Paperspace.
#### sd-concepts-library_Exporter.ipynb
Download all textual inversion embeddings in huggingface.co/sd-concepts-library. May or may not work.
#### Huggingface Textual Inversion Training.ipynb
Train textual inversion embeddings using the HuggingFace API. Doesn't work as well as my `Textual Inversion Training.ipynb` notebook.
#### download_sd-concepts-library.py
Basically `sd-concepts-library_Exporter.ipynb` but a Python script. Can download restricted models.
# Old
#### voldy's_colab_paperspace_09-07-2022.ipynb
First iteration of the WebUI for Paperspace.
#### waifu-diffusion_quick_n_dirty_09-07-2022.ipynb
Quick 'n dirty notebook made to quickly test Waifu Diffusion the day it was released.

View File

@ -1,137 +0,0 @@
import argparse
import datetime
import os
import shutil
import sys
from urllib import request as ulreq
import requests
from huggingface_hub import HfApi
from PIL import ImageFile
def getsizes(uri):
# https://stackoverflow.com/a/37709319
# get file size *and* image size (None if not known)
file = ulreq.urlopen(uri)
size = file.headers.get("content-length")
if size:
size = int(size)
p = ImageFile.Parser()
while True:
data = file.read(1024)
if not data:
break
p.feed(data)
if p.image:
return size, p.image.size
break
file.close()
return (size, None)
parser = argparse.ArgumentParser()
parser.add_argument('out_file', nargs='?', help='file to save to')
args = parser.parse_args()
print('Will save to file:', args.out_file)
# Get list of models under the sd-concepts-library organization
print('Getting list of models...')
api = HfApi()
models_list = []
for model in api.list_models(author="sd-concepts-library"):
models_list.append(model.modelId.replace('sd-concepts-library/', ''))
models_list.sort()
html_struct = """<!DOCTYPE html>
<html lang="en">
<head>
<title>Stable Diffusion Texual Inversion Models</title>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<link href="https://cdn.jsdelivr.net/npm/bootstrap@5.2.1/dist/css/bootstrap.min.css" rel="stylesheet" integrity="sha384-iYQeCzEYFbKjA/T2uDLTpkwGzCiq6soy8tYaI1GyVh/UjpbCx/TYkiZhlZB6+fzT" crossorigin="anonymous">
</head>
<body>
<style>
.img-wrapper {
display: flex;
margin-top: 40px;
}
.img1,
.img2,
.img3 {}
.thumbnail {
height: 185px;
}
.model-link {}
.model-title {
margin-top: 100px;
}
.model-link-wrapper {}
</style>
<div class="container" style="margin-bottom: 180px;">
<div class="jumbotron text-center" style="margin: 45px;"><h1>Stable Diffusion Texual Inversion Models</h1></div>
<p><i>Page updates daily. Last updated {datetime.datetime.now().strftime("%A, %B %d %Y")}.</i></p>
<p>
Generated from <a href="https://huggingface.co/sd-concepts-library">huggingface.co/sd-concepts-library</a>
</p>
<p>
Downloaded models are straight from the HuggingFace repositories and are named learned_embeds.bin. Rename to model_name.pt
</p>
<br><hr>
"""
i = 1
for model_name in models_list:
if i == 3:
break
print(f'{i}/{len(models_list)} -> {model_name}')
# if os.path.exists(f'{model_name}/learned_embeds.bin'): # double check the file exists since sometimes it hasn't been uploaded yet
# shutil.move(f'{model_name}/learned_embeds.bin', f'{model_name}/{model_name}.pt')
# pass
# else:
# continue
# Images can be in a few different formats, figure out which one it's in
img_type = None
img_width = None
for type in ['jpeg', 'png', 'jpg']:
r = requests.head(f'https://huggingface.co/sd-concepts-library/{model_name}/resolve/main/concept_images/0.{type}', allow_redirects=True)
if r.status_code == 200:
img_type = type
img_width = getsizes(f'https://huggingface.co/sd-concepts-library/{model_name}/resolve/main/concept_images/0.{type}')[1][0]
break
html_struct = html_struct + f"""<h3 class="model-title">{model_name}</h3>
<p><a class="model-link" href="https://huggingface.co/sd-concepts-library/{model_name}/resolve/main/learned_embeds.bin">Download {model_name}</a></p>
<p><a class="model-link" href="https://huggingface.co/sd-concepts-library/{model_name}/">View Repository</a></p>
<div class="img-wrapper">
<div class="img1" style="width:{img_width}px;">
<img class="thumbnail" src="https://huggingface.co/sd-concepts-library/{model_name}/resolve/main/concept_images/0.{img_type}">
</div>
<div class="img2" style="width:{img_width}px;">
<img class="thumbnail" src="https://huggingface.co/sd-concepts-library/{model_name}/resolve/main/concept_images/1.{img_type}">
</div>
<div class="img3" style="width:{img_width}px;">
<img class="thumbnail" src="https://huggingface.co/sd-concepts-library/{model_name}/resolve/main/concept_images/2.{img_type}">
</div>
</div>"""
i = i + 1
html_struct = html_struct + '</div></body></html>'
f = open(args.out_file, 'w')
f.write(html_struct)
f.close()